Extended Data Fig. 5: Identification of motor units.

a, Schematic showing three hypothetical, partially overlapping motor units (defined by motor neurons (MNs) B, R and G), tracked over three epochs (i–iii), each characterized by different co-activation patterns (epoch i, R alone; ii, B + R; and iii, B + G). Even though chromatophores 1–4 all belong to the same motor unit (R), their average pairwise correlation during these three epochs would differ owing to the activity of the partially overlapping motor units B and G; identifying motor units using this metric would thus fail. This hypothetical example indicates that the units of coordination during behaviour could be smaller than single anatomical motor units (they could also be larger; for example, if some motor neurons are always centrally coupled). b, Single trials of minimal electrical stimulation experiments in in situ nerves. Composite images (one per trial), green, 10 ms pre-stimulus; magenta, 200 ms post-stimulus; white, overlap. Threshold stimulation either leads to the expansion of a set of three chromatophores (marked with red circles, for example, trial 1), or fails to activate any chromatophore (for example, trial 6, 114 motor units determined with this method). c, Colour assignment of chromatophores in situ. Colour label was assigned based on a threshold on the red channel of RGB space (0.3). Chromatophores (dots) belonging to the same motor unit (as determined in a) are connected by lines, revealing the monochromaticity of motor units. n = 114 chromatophores. d, Dark motor elements tend to be larger than light motor elements. Q–Q plot showing quantiles of the dark versus light motor element size distribution. Line, identity. e, Tail of distribution of motor element spread is heavier with dark than light chromatophores. Q–Q plot showing quantiles of the dark versus light motor element spread (calculated as in Fig. 3d). Line, identity.